[clang] [clang][AMDGPU] Clean-up handling of named barrier type (PR #207687)

[Pierre van Houtryve via cfe-commits]

================
@@ -3,6 +3,30 @@
 // RUN: %clang_cc1 -verify -cl-std=CL2.0 -triple amdgcn-amd-amdhsa -Wno-unused-value %s
 
 void foo() {
+    typedef __amdgpu_named_workgroup_barrier_t SugaredArray[2];
+
+    struct TestSimple {
+        __amdgpu_named_workgroup_barrier_t x;
+    };
+
+    struct TestArray {
+        __amdgpu_named_workgroup_barrier_t y[2];
+    };
+
+    struct TestSugared {
+        SugaredArray z[2];
+    };
+
+    struct GoodWrapper {
+        __amdgpu_named_workgroup_barrier_t x;
----------------
Pierre-vh wrote:

Downstream has a need for it because there is a desire to create wrappers around named barriers with helper functions. Of course the example here is simple but in reality you'd have a few helper functions in a struct like that.

https://github.com/llvm/llvm-project/pull/207687